Feedforward Control
1. Introduction
2. Ratio control
3. Controller design based on steady-
state models
4. Controller design based on dynamic
models
5. Feedback-feedforward control
6. Simulink example
Feedback Control
Advantages
» Corrective action taken regardless of disturbance source
» Minimal process information required for controller design
» PID control is very versatile and usually effective
Disadvantages
» Corrective action not taken until after the output has deviated from the setpoint
» Requires measurement of the controlled output
» Does not allows measured disturbances to be utilized
» Problematic for processes with large time constants and/or long time delays
Feedforward Control
Major disturbance is measured
and used as the controller input
Advantage
» Corrective action can be taken
before the output has deviated from
the setpoint
» Particularly useful for processes
with large time constants and/or
long time delays
Disadvantages
» Disturbance must be measured
» Provides no advantages for other
disturbances
» Typically requires a process model
Ratio Control
Objective is to maintain the ratio of two stream flow rates at a desired value:
» u is the manipulated flow rate
» d is the measured disturbance flow rate
Applications
» Specifying the relative amounts of two components in a blending operation
» Maintaining a stoichiometric ratio of reactants fed to a reactor
» Maintaining a specified reflux ratio in a distillation column
» Holding the fuel-air ratio at an optimal value in a furnace
d
uR
Ratio Control Implementation Methods
Explicit method
» Explicitly compute and
control ratio
» Process gain depends
nonlinearly on disturbance
Implicit method
» Implicitly compute and
control ratio to linearize
process gain
» Ratio station gain:
Sd= span of disturbance stream
flow transmitter
Su = span of manipulated stream
flow transmitter
du
RK
d
p
1
u
ddR
S
SRK
Feedforward Control for Steady-State Models
Distillation column example » Compensate for measured changes in feed flow
rate and feed composition
Steady-state mass balances
Conversion to steady-state feedforward control law » Substitute measurements of disturbances
» Substitute setpoints for controlled variables
» Substitute nominal values from any other variables
xy
xzFD
BxDyFz
BDF
)(
spsp
sp
xy
xtztFtD
])()[()(
Blending System Example
Control problem
» Maintain x at xsp despite measured
disturbances in x1 by manipulating w2
Steady-state mass balances
Feedforward controller
Feedforward controller can provide
setpoint to flow controller
xx
xxww
xwxwwx
www
2
112
2211
21 )(
sp
sp
xx
txxwtw
2
11
2
)]([)(
Feedforward Control for Dynamic Models
Characteristic equation independent of Gf
Ideal feedforward controller
pvt
df
D
Y
pvcm
pvftd
GGG
GG
GGGG
GGGGG
sD
sY
0 :controlPerfect
1)(
)(
pvcm
pvftd
GGGG
GGGGG
sD
sY
1)(
)(
01 pvcm GGGG
Ideal Feedforward Control Examples
Ideal feedforward controller:
Example 1 – implementable
Example 2 – not causal
Example 3 – not proper
pvt
df
GGG
GG
1
1
1
1s
s
KKK
KG
s
KG
KG
KG
s
KG
d
p
pvt
df
p
p
p
vv
tt
d
dd
s
d
p
pvt
df
p
s
p
p
vv
tt
d
dd e
s
s
KKK
KG
s
eKG
KG
KG
s
KG
1
1
1
1
1
)1)(1(
)1)(1(
121
21
s
ss
KKK
KG
ss
KG
KG
KG
s
KG
d
pp
pvt
df
pp
p
p
vv
tt
d
dd
Lead-Lag Feedforward Controllers
Lead-lag unit
» Kf, 1 and 2 are adjustable tuning parameters
Controller tuning
» Controller gain:
» Fine tune Kf to eliminate offset after disturbance
» Controller time constants:
» Fine tune 1 and 2 to improve disturbance rejection performance
1
1
)(
)()(
2
1
s
sK
sD
sUsG ff
pvt
df
KKK
KK
dp 21
1
1
1
1s
s
KKK
KG
s
KG
KG
KG
s
KG
d
p
pvt
df
p
p
p
vv
tt
d
dd
Simulink Example: feedforward_example.mdl
PI controller
Feedforward controller
To Workspace2
setpoint
To Workspace 1
input
To Workspace
output
System1
1
s+1
System
3
5s+1Setpoint
PID Controller
PID
Feedforward
-5s-1
3s+3
Disturbance
Add2Add1
Add
)5/11(3
2)(
15
3)(
5.2 :IMCssG
ssG cp
c
1
15
3
1
)(
)()(
1
1)(
s
s
sG
sGsG
ssG
p
dfd
Setpoint Change
0 2 4 6 8 10 12 14 16 18 20
0
0.2
0.4
0.6
0.8
1
Outp
ut
Time
Output
Setpoint
0 2 4 6 8 10 12 14 16 18 20-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Input
Time